Article Text
Abstract
Objectives To investigate why symptoms indicative of early-stage lung cancer (LC) were not presented to general practitioners (GPs) and how early symptoms might be better elicited within primary care.
Design, setting and participants A qualitative cross-sectional interview study about symptoms and help-seeking in 20 patients from three south England counties, awaiting resection of LC (suspected or histologically confirmed). Analysis drew on principles of discourse analysis and constant comparison to identify processes involved in interpretation and communication about symptoms, and explain non-presentation.
Results Most participants experienced health changes possibly indicative of LC which had not been presented during GP consultations. Symptoms that were episodic, or potentially caused by ageing or lifestyle, were frequently not presented to GPs. In interviews, open questions about health changes/symptoms in general did not elicit these symptoms; they only emerged in response to closed questions detailing specific changes in health. Questions using disease-related labels, for example, pain or breathlessness, were less likely to elicit symptoms than questions that used non-disease terminology, such as aches, discomfort or ‘getting out of breath’. Most participants described themselves as feeling well and were reluctant to associate potentially explained, non-specific or episodic symptoms with LC, even after diagnosis.
Conclusions Patients with early LC are unlikely to present symptoms possibly indicative of LC that they associate with normal processes, when attending primary care before diagnosis. Faced with patients at high LC risk, GPs will need to actively elicit potential LC symptoms not presented by the patient. Closed questions using non-disease terminology might better elicit normalised symptoms.
- Symptoms
- Early cancer diagnosis
- Terminology as topic
- Help seeking
- Lung cancer
- Discourse Analysis
This is an open-access article distributed under the terms of the Creative Commons Attribution Non-commercial License, which permits use, distribution, and reproduction in any medium, provided the original work is properly cited, the use is non commercial and is otherwise in compliance with the license. See: http://creativecommons.org/licenses/by-nc/2.0/ and http://creativecommons.org/licenses/by-nc/2.0/legalcode.
Statistics from Altmetric.com
Article summary
Article focus
-
Why symptoms potentially indicative of lung cancer (LC) are not presented to general practitioners (GPs).
-
Exploration of how and why some LC symptoms are normalised by LC patients.
-
Use of discourse analysis to investigate communication factors involved in the non-presentation and normalisation of symptoms, and how symptoms might be better elicited in primary care.
Key messages
-
Non-specific, episodic and non-progressive symptoms were normalised by patients with operable LC who felt well.
-
Symptoms normalised by patients with operable LC were not presented to GPs during consultations before diagnosis. GP elicitation of normalised symptoms would lead to better-informed referral decisions.
-
Closed questions using non-disease terminology were more effective at eliciting symptoms normalised by patients.
Strengths and limitations of this study
-
This study used interviews to identify interactional factors which influenced symptom presentation within a research study, and it may be that symptom presentation occurs differently within everyday GP consultations; nonetheless our findings indicate that the symptoms normalised by patients within interviews were also the symptoms that consulting patients did not present to GPs. If these normalised symptoms were elicited by GPs, referral decisions would be better informed.
-
Most LC patients are diagnosed with inoperable disease and so any sample of patients diagnosed with operable LC is unrepresentative of this patient population. However, research involving operable patients enables the investigation of communication about currently experienced early symptoms, rather than relying on retrospective accounts of early symptoms provided by patients with later-stage disease. Furthermore, the reasons these patients gave for non-presentation of symptoms concur with other studies of help-seeking for cancer symptoms, supporting the transferability of our findings.
Introduction
Lung cancer (LC) is diagnosed too late in the UK and survival rates are lower than in most other Western European countries;1–4 86% are diagnosed at a stage when curative treatment is not possible and less than 25% survive 1 year following diagnosis.5 ,6 LC kills approximately 30 000 people a year in the UK so even modest improvements in the time to diagnosis could dramatically improve health outcomes.7 Despite successful national cancer screening programmes, most tumours are diagnosed following presentation with symptoms,8 so it is vital to identify patients with significant symptoms early. The UK National Institute of Clinical Excellence (NICE) recommends urgent chest x-ray for patients presenting with any 1 of 10 unexplained or persistent symptoms9 but general practitioners (GPs) have to balance risks associated with unnecessary x-ray against possible late diagnosis, and make judgements about the relative validity of alternative explanations for symptoms. This is further complicated by the fact that LC is often preceded by chronic respiratory disease 10 making detection difficult.
Recent evidence11 indicates that most newly diagnosed LC patients do not recognise all of their cancer symptoms. Isolated single symptoms have low predictive value for LC12 but patients seldom present multiple symptoms to GPs.13 ,14 Interview research has shown that LC patients normalise symptoms and delay seeking help15 ,16 and in the general population many symptoms are never presented to GPs17 ,18 However, patients diagnosed with LC have been shown to report symptoms to their GP more frequently than controls 6–24 months before diagnosis12 but it seems that a combination of cultural and communication processes combine, sometimes fatally, to prevent help-seeking13 ,19 ,20 for the full range of symptoms experienced by patients at an increased risk of LC.11 ,21
Previous studies have identified symptom normalisation—the association of symptoms with normal processes—as an important factor in delayed LC diagnosis. However, research has not yet addressed the reasons for normalisation of LC symptoms, or investigated how normalised symptoms that are not presented to healthcare professionals might be better elicited. Structured interviewing has been used in primary care to improve psychiatric diagnosis but it is not clear if it could help to elicit early LC symptoms. Our study examined how symptoms were normalised by patients and compared structured and unstructured elicitation of symptoms. By using a discourse analytic approach we were able to suggest ways that healthcare professionals might better elicit normalised symptoms, and investigate why they are not presented to GPs.
Methods
Design
Previous studies have focused on inoperable LC, but we were interested in how patients communicated early symptoms so we conducted interviews with patients awaiting surgical resection of LC (suspected or histologically confirmed). Previous interview studies with LC patients have relied upon retrospective accounts of early symptoms experienced before diagnosis. In contrast, we were interested in how patients communicate about, and negotiate the relevance of current early symptoms. In retrospective accounts patients might normalise symptoms to justify delays in seeking help so we also investigated the normalisation of symptoms that started following LC investigation. We used unstructured followed by structured interviewing to find out if this could elicit symptoms more effectively than open questions about changes in health, which have been found not to elicit all LC symptoms.11
Participants
The interview sample for this study was drawn from 28 adult patients with a diagnosis of, or suspected of having, operable LC (probable: >90% or histologically confirmed) recruited to a questionnaire development study. Patients were either approached by the researcher following their first consultation with participating thoracic surgeons at a South England Trust, or were sent a letter and information sheet by the surgical team. Seventeen of 20 consecutive patients within three recruitment periods (07/2006–10/2007; 02/2008–05/2008 and 02/2009–05/2009) approached by a researcher agreed to take part. An opportunistic sample of 11 participants was recruited by letter (within the three recruitment periods). Twenty-eight patients in total were recruited and interviewed about their current and recent health and help-seeking behaviour.
This paper reports the analysis of 20 interviews with patients identified as having operable LC at the end of the study period (data from seven interviewees who received a non-malignant diagnosis after the interview were analysed separately and are not reported here. One patient diagnosed with advanced disease was also excluded). Characteristics of these 20 patients are given in table 1.
Interviews
The unstructured (first) section of the interview used open questions to generate narrative accounts of participants’ experiences and changes in health status (see online supplementary appendix S1 for the interview checklist). Participants were asked to describe anything at all that they had noticed about their health, even if they thought it not relevant to their investigation for LC. The second part of the interview was semistructured and focused on the duration and characteristics of symptoms, and reasons for seeking or not seeking help. The third part of the interview used closed questions to explore symptoms and help-seeking using a list of potential LC symptoms compiled from Cancer Research UK6 information, NICE9 guidelines and a previous interview study with LC patients.15 Field notes were recorded after the interview. Interviews lasted between 1 and 2 h, took place in the participants’ home (18/20) or a hospital setting (2), some involved the participant's partner (2) or carer (1), all were audio-recorded, transcribed verbatim, checked for accuracy and anonymised. An adapted version of Jefferson's transcription conventions22 were used (described in box 1).
Transcription notation Simplified and adapted version of jeffersonian transcribing conventions
-
The speaker is identified by a participant identifier (P1–P28) followed by a colon. The participant's partner is indicated by a P following the participant identifier for example:
P24P: No I do not agree
-
Round brackets indicate that the material in the brackets is either inaudible, for example:
-
M: I () that
-
Or there is doubt about its accuracy, for example:
M: I (could not tell you) that
-
A micropause (a noticeable pause of less than 0.2 s) is indicated by a dot enclosed in brackets: (.)
-
Non-verbal activities and noticeable pauses of 0.2 s or more are indicated within double brackets:
M: Yes ((laughter)) but ((pause)) I do not know
-
Square brackets indicate that material has been removed, usually to protect the participant's identity, for example:
[ ] or [town]
-
Three consecutive dots indicates that a section of transcript has been removed:
M: He ran up the hill…to the house at the top
-
Square brackets between adjacent lines of speech mark the start and end of overlapping talk [ ]
Analysis
The first stage of analysis involved an iterative coding process using elements of the constant comparative method to develop themes (initially identified by LB and checked by a second researcher, GL, who independently read a sample of transcripts and verified codes and themes). This iterative process continued until data saturation was achieved. All transcripts were revisited and deviant cases were sought.23 Thematic analyses identified symptoms not presented to GPs, characteristics of symptoms and reasons given for non-presentation. Discourse analysis24 ,25 which considers language use in context, was used to examine how health changes were presented in patient–interviewer interactions; the discourse analysis was informed by ethnomethodology, an approach which focuses on how social action is accomplished within accounts. This enabled us to look at the implications of talks sequential and microorganisation for symptom presentation, and showed how normalised symptoms might be better elicited. We combined the thematic analysis and discourse analyses to explain normalisation and non-presentation of symptoms. The results section presents key findings about symptom presentation, including reasons for non-presentation, and the implications of question type and terminology.
Results
Most participants described themselves as having good health; only four presented accounts of declining health preceding diagnosis, characterised by multiple symptoms and feeling unwell (see table 2).
Symptomatic diagnosis occurred for 13 participants and 7 participants claimed not to have any LC symptoms, describing incidental diagnoses made during the investigation of unrelated health problems, traumatic injury or screening (table 3).
Fifteen participants described further changes in health possibly indicative of LC (according to NICE guidelines/CRUK symptom list) that were not thought a reason for concern and had not been presented to their GP during LC investigations, despite the presentation of the trigger symptom or use of primary care services for other reasons. They did not associate these uninvestigated health changes with LC and they were elicited by closed questions about specific symptoms, but not by open questions about symptoms or changes in health (table 3).
Two types of symptom accounts were identified: ‘symptoms as normal processes’ and ‘symptoms of disease/concern’. Examples of these accounts and their elicitation are provided in table 4. Participants reported uninvestigated symptoms, and produced normalised accounts of these, irrespective of patient sociodemographic characteristics, smoking status or route to diagnosis; there were no discernible differences in relation to table 1 characteristics. Exceptions appeared to arise only in the case of participants providing narratives of declining health. The association of symptom normalisation with narratives of good health is highlighted in table 3; those providing narratives of declining health tended not to normalise symptoms. Participants with incidental diagnoses also provided normalised accounts of uninvestigated potential LC symptoms, but were less likely to produce symptoms of concern accounts than those with symptomatic diagnoses (see table 3).
The first results section—‘Reasons for non-presentation’—describes the main features of ‘symptoms as normal processes’ accounts (episodic/non-progressive symptoms or ageing and lifestyle-related explanations). ‘Symptoms of concern’ accounts are described in order to demonstrate exceptions to the normalisation of symptoms. The second results section examines the use of closed questions to elicit (normalised) accounts of symptoms not elicited by open questions or presented to GPs, and the implications of symptom terminology.
Reasons for non-presentation
Normal processes such as lifestyle and ageing were commonly used as explanations for not presenting symptoms to GPs. For example, breathlessness was frequently associated with being unfit, getting older, overactivity or seasonal changes rather than LC:
P18: I just put it down to me being too unfit for that particular run or circuit or down to age…I didn't associate that with anything other than me being old or unfit, one of those.
In these ‘symptoms as normal process’ accounts patients portrayed symptoms as part of everyday life processes and avoided claiming cancer causation:
LB: ..do you get any discomfort anywhere, do you have any aches or pains?
P11: No (.) only round me neck but that's just recently it's come on. I don't know whether it's to do with this problem I've got...I think it's a bit of arthritis there. And (.) you know (.) it's old age really I mean, because we do get these things I know, as you get older, (.) but just as I say this last couple of weeks it's got really really bad.
Some of those who described current ‘good health’ at odds with their diagnosis, also described episodic ill health, or long-term symptoms which had led to lifestyle changes and adaptation. Symptoms like breathlessness or cough might be more severe during a chest infection, but were not commented on if they persisted. Here, P25 did not mention breathlessness on climbing the stairs to her GP:
P25: It was getting the pains in my hands and my wrists… It was when it started here [in wrists], it started to hinder me with things…but I wasn't going [to the GP] through breathlessness …because that had finished when I got better…You know within the week I was back to being able to breathe again. Apart from when I you know whether you get out of breath carrying the hoover upstairs...[Husband] says “What have you been doing? [ ]?” and I just say “Nothing just those stairs”.
The ability to improve did not appear to fit with the expected progressive pattern for a disease such as LC:
P25: ..[ ].once I've had my antibiotics or a bit of an inhaler I'm fine again, like I am now...Why don't I feel really, really ill now to understand this? How can you have this and get better and feel better, get ill but then you get better, well how can you do that?
These normalised accounts, by simultaneously presenting alternative non-disease explanations, such as ageing, for health changes, also helped construct the participant as healthy. Exceptions to the use of normalised accounts for uninvestigated symptoms were found in four interviews where patients had declining health (consisting of multiple symptoms and feeling unwell); two of these four patients also provided ‘quest for diagnosis narratives’ in which they had battled, or were still battling, for a diagnosis in the face of clinical ignorance or clinical delay. In the interviews they described most of their health changes in response to open questions (table 3: exceptions to the normalisation of symptoms), including symptoms not presented to GPs, and did not normalise these symptoms. Even symptoms presented in response to closed questions were most often not normalised:
LB: So have you noticed any changes in breathing or breathlessness?
P19: Yes I am definitely more breathless now...I am not normally that breathless!
LB: ...and before that, how would you describe your breathlessness?
P19: Well it's never been really too bad, as long as I've had my inhalers...So it's just recently that I am beginning to get a bit more breathless and I don't think that's associated with the asthma.
Participants who presented themselves as well, normalised non-specific, non-progressive and episodic symptoms.
Using closed questions to elicit symptoms not elicited by open questions
Symptoms interpreted as normal by participants tended not to be described in response to open interview questions (tables 3 and 4) and were not presented to GPs. For example, P22, who had been investigated by his GP for a bowel disorder in the weeks before diagnosis, described an absence of symptoms he associated with LC:
P22: No as I say this was a complete shock to find out that it was on the lung. As I said, we would never have known anything about it if I hadn't fallen off that thing. I suppose it would eventually with finding this I suppose I could have lost weight or gone awful thing one to the doctor “well we'll have to find out what's causing it” but no nothing.
However, when asked specifically about long-term cough, he revealed that he had experienced a cough for 4–5 months:
P22: Well I've got a cough now. Every now and again I cough and get a little phlegm up.
LB: … And is it something that you ever went to your doctor about?
P22: No.
LB: No.
P22: No I've never had to do that.
Accounts produced in response to closed interview questions about specific symptoms displayed two common structures for symptom reporting: ‘Affirmation/Normalisation’ and ‘Delayed Affirmation/Normalisation’. The symptom referred to by the interviewer might either be affirmed but normalised (‘Affirmation/Normalisation’) or initially denied and then normalised (‘Delayed Affirmation/Normalisation’). When closed questions phrased health changes in ways which did not necessarily indicate disease, the participant was more likely to answer affirmatively, or describe a health change, but then suggest the symptom was normal and not related to LC (Affirmation/Normalisation). In contrast, questions using disease-related terms—for example, ‘pain’—produced an immediate denial or pause (non-response) followed by normalisation (delayed affirmation/normalisation):
LB: ...have you had any chest pain at all that you can describe?
P12: No, not really. I mean as the cough's got shall we say more persistent and sort of shall we say worse yes (.) I can feel it a bit (.) but I mean I can't feel it now...if you look at the x-rays you think ‘oh blimey!’ but you wouldn't know it was there!
Reformulation of the question, involving a shift from disease to non-disease terminology, could elicit normalised accounts of symptoms—as in these examples where a change in terminology shifting from ‘pain’ to ‘aches’ and ‘discomfort’, and shifting from ‘breathlessness’ to ‘not being able to get your breath’ leads to elicitation of the symptoms:
Excerpt 1:
LR: Have you had any pain anywhere?
P16: None at all. No
LR: …have you experienced any sort of aches or general sort of discomfort at all? ...
P16: No, not serious no. Well …sometimes I have a feeling that something's going on, but it's not there all the time, you know
Excerpt 2:
LR: And have you experienced breathlessness? ((pause))
P18: ((intake of breath))
LR: Just feeling like you haven't been able to get your breath quite so easily?
P18: I would go up a couple of flights of stairs quite randomly, I would feel out of breath. I would never never usually be like that, so yes, for a fit guy I would go ooh I'm breathless...but then you know I shouldn't have really bothered about it at all. But then again I have put on a slight bit of weight haven't I?
In contrast to disease-related terminology, terminology not strongly associated with disease such as ‘aches’ or ‘discomfort’ rather than ‘pain’ produced affirmation and then normalisation (affirmation/normalisation):
LB: And have you had any kind of aches or discomforts anywhere?
P12: Well I have been complaining about a stiff neck haven't I...and also this shoulder…but I mean I can play golf, so it's not that bad!
Similarly, the use of terms that imply ‘breathing changes’ or ‘getting out of breath more easily’, rather than ‘breathlessness’, produced an affirmation/normalisation response structure:
LB: ...what about breathing changes, or have you ever noticed at all that you can become more breathless than you would have done say a few years ago when you were doing something?
P11: I do now. This past (.) oh couple of months I suppose. I get more breathless if I (.) if I hurry around too much you know…but normally you know, I don't run around! (LB: no no) If I remember my age...I don't sort of get out of (.) breathless or anything like that, it's only if I'm (.)...overdo things really.
Even though closed questions using disease-related terminology might elicit previously unmentioned health changes, closed questions using non-disease terminology did so more effectively.
Discussion
Eliciting ‘hidden’ symptoms
Most of our sample described themselves as feeling well, despite going on to have a diagnosis of operable LC. Patients who felt well had experienced a range of health changes indicative of LC but they did not tell their GPs about many of these, despite making use of primary care services. Instead they framed these symptoms as normal features of lifestyle and ageing processes.
Delay in LC diagnosis in the UK has been blamed upon patients’ failure to recognise early symptoms.26 However, our research indicates that normalised symptoms can be elicited by closed questions. This runs counter to current educational and communication practice which encourages open and expansive questioning. Whereas open questioning is necessary to elicit symptoms perceived as abnormal by the patient, normalised symptoms will remain hidden. Once elicited by closed questions, normalised symptoms are often quickly obscured within accounts which provide every day explanations for health changes. This means that interviewers (or health professionals) have to probe normalised accounts to uncover hidden symptoms.
Questions using disease-related symptom terminology, such as ‘chest pain’, or ‘breathlessness’, appeared to have limited potential to elicit potential LC symptoms experienced by those with operable LC. Our analysis suggests that to get at these symptoms we need to ask closed questions without referencing disease-related symptom labels. Again this runs counter to some guidance such as the NICE referral criteria terminology which uses disease-related terms. Furthermore, contextual factors and framing of the patient's presentation are known to influence GPs’ diagnostic reasoning;27 patients who present themselves as well and without declining health might less likely raise concern and be referred for an investigation of potential cancer symptoms.
Recent survey research looking at public awareness of cancer symptoms in the UK concluded that levels of knowledge are low for many potential cancer symptoms.28 These findings have informed regional National Awareness and Early Diagnosis Initiatives29 materials designed to educate the public about cancer symptoms. It may be argued that participants in our study simply did not recognise the significance of symptoms such as breathlessness. However, participants did not report lack of knowledge as the reason for symptom normalisation and non-presentation. Furthermore, the accounts produced by participants avoided personal claims of LC aetiology for changes in health, even if this was raised as a possibility in the interview. Alternative non-LC explanations for symptoms were provided that had social legitimacy. Our work suggests that lists of symptoms alone are unlikely to prompt patients to reveal multiple non-specific and normalised symptoms, especially when they are asked to give unstructured accounts. Furthermore, our research indicates that LC risk scores provided by symptom-based clinical decision support aids (eg, RATS30) are likely to be influenced by how symptoms are elicited within the consultation.
Patient-centred medicine attempts to honour patients’ experiences and concerns—presented in their own terms. It has been accompanied by more open consultation styles and a shift away from interactions directed by the health professional. For patients with potential LC this may not be the best way to elicit symptoms. Instead routine medical consultations involving those at increased cancer risk31 might better be restructured to enable the presentation of health changes which appear normal or unproblematic to the patient. This would require the clinician to be aware of the risk of LC in all patients presenting to their service with symptoms seemingly unrelated to LC. The elicitation of normalised symptoms in patients at increased LC risk might then facilitate GPs’ chest x-ray referral decisions.
Strengths and limitations
This study used interviews to identify interactional factors that influenced symptom presentation within a research study. The systematic and in-depth study of language of the type reported in this article can lead to critical insights about conventions used in conversation that are transferable between settings.32 However, it may be that symptom presentation occurs differently within everyday GP consultations; closed questions involving non-disease terminology might not be as effective at eliciting normalised symptoms within primary care practice. Further research involving GP consultations will be required to establish how effective these methods of symptom elicitation are within primary care. Nonetheless, our findings indicate that the symptoms normalised by patients within interviews were also the symptoms that consulting patients did not present to GPs. If these normalised symptoms that are potentially indicative of LC were elicited by GPs, referral decisions would be better informed.
The participant group included patients with an established or probable LC diagnosis. This may influence symptom presentation in the interview as an LC diagnosis is already suspected. However, the normalisation of symptoms that started after diagnosis within this study suggests that normalisation is not justifying delays in diagnosis; the association of episodic, non-specific symptoms with normal processes appears commonplace for those feeling well, even when LC provides a potential explanation for symptoms.
NICE referrals guidelines for suspected LC are based upon a weak evidence base; therefore, we do not know the likelihood that the symptoms not presented to GPs were caused by LC. However, these guidelines represent the best evidence currently available to inform referral for LC investigation. If these non-specific symptoms experienced by patients at increased LC risk were elicited in primary care, GPs would be better able to operationalise NICE guidelines. A prospective study may eventually determine the utility of these symptoms in the early diagnosis of LC and the efficacy of treatment (including surgery).
The majority of LC patients are diagnosed with inoperable disease and so any sample of patients diagnosed with operable LC is necessarily unrepresentative of the whole population of LC patients. It may be that our participants were more symptomatic in the early stages, or more likely to seek medical help, than those diagnosed with inoperable disease. However, this makes it all the more compelling that these participants still experienced a number of symptoms that they did not report. The reasons these patients with LC give for non-presentation of symptoms concur with other studies of help-seeking for cancer symptoms,19 supporting the transferability of our findings. Furthermore, our finding that those who reported good health tended to normalise non-specific, episodic and non-progressive symptoms might have implications for improving the earlier detection of other cancers where patients describe good health in the early stages and for patient–clinician communication more generally.
Conclusions
Even though LC patients are more likely to attend their GP with potential symptoms in the year before diagnosis than healthy controls, our findings indicate that many non-specific symptoms are not presented within these consultations. The use of non-disease related symptom labels in combination with some closed questioning appears to improve symptom elicitation.
Eliciting and investigating normalised symptoms—to uncover the invisible part of the illness iceberg,17 ,18 while not feasible for all patients attending primary care, would be possible for those identified as at increased LC risk.31
Acknowledgments
Liz Roffe conducted some interviews. The structured section of the interview was based upon an interview schedule developed by Corner, Hopkinson, Fitzsimmons et al. (2005).
References
Supplementary materials
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Files in this Data Supplement:
- Data supplement 1 - Online supplement
- Data supplement 2 - Online supplement
Footnotes
-
Contributors All authors have commented on the first and final draft of the paper, and contributed to interpretation of the data. LB was the chief investigator; she had the original idea, designed the study, conducted interviews, analysed the data and produced all drafts. CP also commented on the penultimate draft of the paper and edited the final draft. GL contributed to coding of the interview transcripts. AB also contributed to data collection. LB is the guarantor for the study.
-
Funding This research was funded by an Research Councils UK post-doctoral fellowship awarded to LB to develop research to improve earlier lung cancer diagnosis (sponsor: Faculty of Health Sciences, University of Southampton—Jessica Corner).
-
Competing interests None.
-
Ethical approval NHS ethical approval was gained from Southampton South Central Research Ethics Committee (05/Q1702/46).
-
Provenance and peer review Not commissioned; externally peer reviewed.
-
Data sharing statement Research participants have not given informed consent for data sharing; a complete qualitative data set cannot be anonymised adequately.